MABQ BigQuery Agent
MABQ is an AI-powered agent built with Google’s Agent Development Kit (ADK) and Gemini 2.5 Pro that translates natural language questions into SQL queries for BigQuery. Designed for enterprise deployment, it combines powerful AI capabilities with robust security controls.What is MABQ?
MABQ (short for “Modelo Asistencial BigQuery”) is a conversational AI assistant that allows users to query BigQuery datasets using natural language. Instead of writing SQL, users can ask questions like “Show me the top 10 assets by value” and receive accurate query results. The system consists of:- Backend Agent: Built with Google ADK, powered by Gemini 2.5 Pro LLM
- Frontend Interface: Next.js application with CopilotKit integration
- Microsoft Teams Integration: Deploy as a Teams app with SSO authentication
- Enterprise Security: Azure AD authentication and read-only BigQuery access
Key Features
Natural Language to SQL
Convert plain language questions to BigQuery SQL using Gemini 2.5 Pro
Enterprise Security
Azure AD JWT authentication with read-only database controls
Teams Integration
Deploy directly in Microsoft Teams for seamless enterprise adoption
Cloud Native
Serverless architecture on Google Cloud Run with auto-scaling
Architecture Highlights
- Google ADK Framework: Built on Google’s Agent Development Kit for production-grade agent orchestration
- Vertex AI Integration: Direct access to Gemini 2.5 Pro through Vertex AI
- BigQuery Toolset: Specialized tools for schema inspection and query execution
- Security Guardrails: Prevents data modification with read-only mode enforcement
- FastAPI Backend: High-performance async API with CORS and authentication middleware
- CopilotKit Frontend: Modern chat interface with streaming responses
Use Cases
MABQ enables business users, analysts, and developers to:- Explore data without SQL knowledge
- Generate reports through conversational queries
- Validate datasets with natural language questions
- Accelerate analytics by reducing time from question to insight
Getting Started
Quickstart
Get MABQ running locally in 5 minutes
Architecture
Understand the system design and components
Agent Development
Learn how to configure and customize the agent
Deployment
Deploy to Google Cloud Run for production
Technology Stack
- Agent Framework: Google ADK (Agent Development Kit)
- LLM: Gemini 2.5 Pro via Vertex AI
- Backend: FastAPI (Python 3.11)
- Frontend: Next.js 16, React 19, CopilotKit
- Data Layer: BigQuery
- Authentication: Azure AD (Microsoft Entra ID)
- Deployment: Google Cloud Run (containerized)
- Integration: Microsoft Teams SDK
Next Steps
Run the Quickstart
Follow the quickstart guide to run MABQ locally and test natural language queries
Explore the Architecture
Review the architecture documentation to understand how components interact
Configure Your Agent
Customize the agent’s behavior and BigQuery access in Agent Configuration
Deploy to Production
Deploy to Google Cloud Run following the deployment guide